SDN中基于分布式决策的控制器负载均衡机制

针对SDN多控制器负载均衡过程中,控制器选取僵化和交换机迁移冲突问题,提出了一种基于分布式决策的控制器负载均衡机制,分为三个阶段进行实施：首先通过周期性收集网络信息,结合控制器负载状况构建分布式迁移决策域;然后在域中依据选取概率确定迁移交换机,综合权衡数据收集、交换机迁移和状态同步三种代价选择目标控制器;最后建立迁移时钟模型,完成交换机迁移和控制器角色转换.仿真结果表明,与现有的负载均衡机制相比,降低了网络的通信开销,流建立时间平均缩短0.14s,控制器资源利用率提高了21.7%.     

SDN中基于Q-learning的动态交换机迁移算法

由于网络流量动态变化,控制器负载均衡成为大规模部署软件定义网络研究的重点.提出基于Q-learning的动态交换机迁移算法,首先对软件定义网络中的控制器部署问题建模,再应用Q-learning反馈机制学习实时网络流量,最后根据Q表格将交换机从高负载控制器动态迁移到低负载控制器上,实现控制器的负载均衡.仿真结果表明,所提算法能够获得较低的控制器负载标准方差.     

SDN中基于效能优化的交换机动态迁移策略

交换机迁移作为一种弹性控制方式对于软件定义网络（SDN）中多控制器负载均衡具有重要作用.针对现有迁移方案存在迁移效率低下和高迁移代价问题，提出基于效能优化的交换机动态迁移策略.首先，分析控制器负载组成并构建负载差异矩阵，设置触发因子进行负载不均衡判定.然后确定迁移对象，建立迁移效能模型，同时考虑负载均衡率和迁移代价来确定迁移交换机和迁入控制器.最后，通过设定迁移三元组完成迁移映射，实现高效交换机迁移.仿真结果表明，该策略可以有效降低控制器响应时间，减小迁移代价并提高控制器吞吐量，同时将负载均衡率保持在较高水平，具有良好的拓扑适应性.     

多域SDN网络中负载均衡路由算法研究

SDN是一种新型的网络架构,可分离数据平面和控制平面。其通常利用集中控制器来管理所有的网络中的交换机。随着网络规模的扩大,单个控制器的有限性能将导致控制平面拥塞,一些方案被提出以解决可扩展性问题,即将网络分离多个域。文章分析了现有的跨域交互技术以及负载均衡算法,并对已有的负载均衡算法作出比较。     

软件定义网络中基于效率区间的负载均衡在线优化算法

在大型复杂软件定义网络中,为提高网络负载均衡,减少控制器与交换机间的传播时延,该文提出一种基于效率区间的负载均衡在线优化算法。在初始静态网络中,通过贪心算法选择初始控制器集合,并以其为根节点构建M棵改进代价的最小生成树(MST),确定初始M个负载均衡的子网;当网络流量发生变化时,通过广度优先搜索(BFS)调整子网间交换机映射关系使其满足效率区间,保证任意时刻网络的负载均衡。算法均以网络连通性为基础,且均以传播时延为目标重新更新控制器集合。仿真实验表明,该算法在保证任意时刻网络负载均衡的同时,可以保证较低的传播时延,与Pareto模拟退火算法、改进的K-Means算法等相比,可以使网络负载均衡情况平均提高40.65%。     

基于流量工程的软件定义网络控制资源优化机制

针对软件定义网络(SDN)分布式控制平面中由于网络分域管理所引发的控制扩张问题,该文提出了一种基于流量工程的SDN控制资源优化(TERO)机制。首先基于数据流的路径特征对流请求的控制资源消耗进行分析,指出通过调整控制器和交换机的关联关系可以降低控制资源消耗。然后将控制器关联过程分为两个阶段:先设计了最小集合覆盖算法来快速求解大规模网络中控制器关联问题;在此基础上,引入联合博弈策略来优化控制器和交换机的关联关系以减少控制资源消耗和控制流量开销。仿真结果表明,与现有的控制器和交换机就近关联机制相比,该文机制能在保证较低控制流量开销的前提下,节省约28%的控制资源消耗。     

基于控制器集群的SDON多域互联互通技术北大核心

针对软件定义光网络(SDON)多控制域互联互通的需求,在对多控制域信息交互协同控制进行分析研究的基础上,利用负载均衡的思想,提出了一种基于控制器集群的SDON多控制域互联互通架构,包含集群间和集群内两个互联层次。利用集群间控制器迁移和集群内主从控制器切换机制,实现光网络中多控制域间的协同互联和负载均衡,提升网络服务质量。     

软件定义网络中基于负载均衡的多控制器部署算法

随着软件定义网络规模扩大,控制层与数据层解耦带来了诸如控制器部署等新问题。该文提出基于负载均衡的多控制器部署算法(Multi-Controller Deployment Algorithm Based on Load Balance, MCDALB)。算法首先根据网络拓扑结构及其负载情况,确定控制器数量K;然后根据控制器容量限制,提出一种近似比为2的多控制器负载均衡算法,将网络划分成K个控制区域;最后根据区域内所有交换机到控制器距离总和最小原则,在控制区域部署控制器。为了验证算法的性能,选取实际网络拓扑进行实验。实验结果表明,与AL, WL算法相比,该算法在满足控制器负载近似比为2的同时,网络最大延时差距不超过0.65 ms。     

软件定义网络中基于时延和负载的多控制器部署策略研究

多控制器体系结构的出现,解决了经典软件定义网络(SDN)架构控制层以单一集中控制器为主,在大规模网络环境中的可扩展性问题.在多控制器体系结构中,由于生成转发规则并将其填充到交换机的任务被委托给了控制器,网络的性能在很大程度上取决于控制器的放置.该文以降低总时延和均衡控制器间负载为目标,提出了一种基于子网划分的多控制器部...     

软件定义网络中面向时延和负载的多控制器放置策略

在多控制器管理的软件定义网络(SDN)中,时延和负载是控制器放置问题(CPP)要考虑的重要因素。该文以降低控制器之间的传播时延、流请求的传播时延和排队时延、均衡控制器间负载为目标,提出一种控制器放置及动态调整的策略,其中包括用于初始控制器放置的负载均衡算法(BCRA)和遗传算法(GA),用于动态调整控制器负载的在线调整算法(ADOA)。以上算法均考虑网络连通性。仿真结果表明:在初始控制器放置时,在保证流请求的传播时延、排队时延和控制器传播时延较低的情况下,BCRA部署在中小型网络中时,其负载均衡性能与GA相近且优于k-center和k-means算法;GA部署在大型网络中时,与BCRA, k-center和k-means算法相比,使得负载均衡率平均提高了49.7%。在动态情况下,与现有动态调整算法相比,ADOA可以保证较低排队时延和运行时间的同时,仍能使负载均衡参数小于1.54。     

基于备份的可重构服务承载网可靠性映射方法

可重构柔性网络链路失效将严重影响其上承载的可重构服务承载网(RSCN)的可靠性。文章基于路径备份策略着重解决时延敏感类型RSCN的可靠性问题,并提出分阶段处理方式进一步优化备份资源消耗。在拓扑预处理阶段,根据RSCN是否支持路径分裂分别提出分裂的最小备份拓扑生成(S-MBT-Gen)算法和最小备份生成树(MBST-Gen)算法,减小备份拓扑带宽约束总量;在拓扑映射阶段,提出主备拓扑协同映射(RNM-PBT)算法,协调利用底层网络资源。仿真结果表明,本文提出的算法降低了RSCN平均资源消耗,且具有较高的请求接受率和较低的平均执行时间。     

运用备份服务位置和概率QoS模型的Web服务组合算法

针对工作流服务数的增加过程中最初规划的潜在成本较高以及很多服务组合算法可靠性不足等问题,提出了一种利用备份服务位置和概率服务质量（QoS）模型的服务组合算法。该算法计算服务集群的QoS优化选择,为每个服务包含足够数量的备份服务,且考虑了备份服务位置,以便在单个故障点上进行评估。由于这些备份服务分布均匀,防止了任务失败的发生。对于服务选择问题,采用一种改进的多目标优化（MOO）算法,利用聚类和QoS模型来计算可行解集合。仿真实验采用JMETAL 3.1框架,评估备份服务位置的收益以及算法的可靠性。结果表明,相比于其他MOO算法,提出的算法可靠性更高,从备份服务位置所获得的收益更高。     

Research on switch migration method based on minimum cost path

In order to protect the controller,especially the controller in backbone network,from security threats and attacks,improve the security of the software-defined network (SDN) control plane,a switch migration algorithm based on minimum cost path was proposed.A load prediction module was added to the migration model,which executed a controller load prediction algorithm to obtain a load prediction matrix,and then a migration-target controller set was determined according to the load prediction matrix.The improved Dijkstra algorithm was used to determine the minimum cost path.According to the load state of the controller and the traffic priority of the switch to be migrated,the optimal migration switch set was determined.The problem of isolated nodes was solved that may occur during the migration process.The experimental results show that the migration timing of the algorithm is more reasonable,the selection of the migration controller and the target controller is more reasonable,the load balancing of the control plane is realized,the number of migrations and cost are reduced,and the performance of the controller is improved.     

支持文件重删的HDFS分布式备份系统研究

随着传统行业电子化的速度日渐加快,电脑上的数据量与日俱增,数据备份领域面临的挑战也越来越大.传统备份因为需要大量的磁盘阵列作为存储介质,所以在成本控制上一直是个难题.针对该情况,结合云存储平台低廉的成本、高效的资源伸缩与利用率,研究在HDFS系统中加入新的重复数据删除技术,并对原有备份策略进行优化,设计出一种基于云存储中的重复数据删除技术的备份系统.最后通过实验,对改进后的方案系统与传统备份方案备份文件所占空间、时间等参数进行了对比.     

一种基于OPM200和IBBS50L的极简基站

在5G网络、大数据中心、边缘计算等新基建快速推进的大背景下,受限于存量天面空间、存量供备电系统、机房空间容量不足以及站址租金居高不下等问题,5G网络的大规模建设给通信公司运营成本方面带来巨大挑战。文章从引入背景、改造方案及成效等方面入手,探讨基于华为刀片电源OPM200作为主电、IBBS50L锂电作为备电的极简基站如何破解5G发展遇到的难题。     

Per-user checkpointing for mobility database failure restoration

This paper studies the failure restoration of mobility database for Universal Mobile Telecommunications System (UMTS). We consider a per-user checkpointing approach for the home location register (HLR) database. In this approach, individual HLR records are saved into a backup database from time to time. When a failure occurs, the backup record is restored back to the mobility database. We first describe a commonly used basic checkpoint algorithm. Then, we propose a new checkpoint algorithm. An analytic model is developed to compare these two algorithms in terms of the checkpoint cost and the probability that an HLR backup record is obsolete. This analytic model is validated against simulation experiments. Numerical examples indicate that our new algorithm may significantly outperform the basic algorithm in terms of both performance measures.     

Optimized design and availability analysis of large-scale shared backup path protected networks

Traditional integer linear programming model for shared backup path networks allows only one working route and one backup route per demand and does not scale well. By introducing multiple working routes and backup routes, the traditional multi-flow model solves in a faster manner. This paper seeks improvements on the traditional multi-flow model and develops an algorithm to assess availability for multi-flow shared backup path protection models. Experiments on 165 networks testify that the newly proposed model is 51% faster on average with similar total cost and overall network availability, compared with the traditional multi-flow model. All the networks in this paper are designed to be 100% single-failure restorable, and major findings regarding these networks include: (1) total cost of assigning backup capacity to each span dwindles away with increasing network average nodal degree; (2) network availability first rises then falls as network average nodal degree increases; and (3) when network scale increases, network availability decreases with fluctuations. The results found are explained with two case studies in this paper.     

Personal Cloud P2P

Small and medium sized businesses have lacked of good and automate backup system. Data storages are unreliable and tend to fail without any warning. Whereas, personal computers are already equipped with terabytes of hard drives. This is enormous space for personal or business use. Under normal working conditions, more than half of total storage space is not being used. It is scattered in all connected devices. This paper proposes pooling unused storage resources to create a personal private cloud. The process will be transparent to the users by having a reliable drive with an auto backup built in. The proposed system solution uses a replica-based model in which three sets of the same information would distribute and keep in three different nodes. All the participated nodes are peer-to-peer (P2P) and are decentralized. Therefore, private cloud system can support small or medium businesses as backup system or business sharing drives without adding any cost.

     

Three-dimensional scan-based algorithm for directional neighbor discovery in ad hoc networks

In recent years, neighbor discovery techniques using directional antennas have attracted widespread attention. Most currently available directional neighbor discovery techniques are designed based on two-dimensional (2D) space. Although these algorithms can accomplish the discovery between nodes, the algorithm portability is poor for three-dimensional (3D) platforms, and the application scenarios are limited since new communication scenarios such as unmanned aerial vehicle (UAV) groups and maritime fleets emerge where information needs to be delivered in real time and nodes are located in 3D space. This paper proposes a new deterministic directional neighbor discovery algorithm in 3D space named 3D scan-based algorithm (3D-SBA) to meet the needs of the line-of-sight (LOS) scenes. Four existing neighbor discovery algorithms, namely, SBA-based random mode selection (SBA-R), quorum, complete random algorithm (CRA), and SBA-based leader election algorithm (LE), have been extended to our proposed 3D algorithm model for simulation and comparative analysis with 3D-SBA. The simulation results show that the 3D-SBA algorithm consumes 51.15%, 180.20%, 9.48%, and 17.49% of the time slots compared with the four existing algorithms mentioned above. However, the quorum algorithm has a very high node collision rate, up to 92.11%. Ultimately, the 3D-SBA algorithm has the best performance considering the conflicts and the density of network nodes.